Medicine (RMH) - Theses

Permanent URI for this collection

http://hdl.handle.net/11343/194

Filters

2022 1
1
Reset filters

Settings
Statistics
Citations
Type: Doctorate × Author: Chong, Simon Woon-Hui ×

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Publication bias and the effect of mandatory trial registration in the anaesthesia literature
    Chong, Simon Woon-Hui ( 2022)
    Publication bias and selective outcome reporting have been demonstrated in many areas of medicine, and can have significant impact on clinical practice through the weighting of published evidence toward selected positive results. Publication bias and selective outcome reporting can influence the findings of meta-analyses and summaries of evidence, and are significant contributors to the substantial problem of increasing waste in research. The aim of this thesis was to investigate publication bias in the anaesthesia literature and the effect of the introduction of mandatory prospective trial registration on this bias. In Chapter 2 we attempted to quantify the amount of publication bias present in the anaesthesia literature in the era prior to introduction of mandatory prospective trial registration. We conducted a review of all randomised controlled trial (RCT) abstracts presented at the American Society of Anesthesiologists (ASA) annual meetings between 2001 to 2004. By assessing whether the primary outcome of the abstract was positive or negative and then eliciting any subsequent journal publication of the abstract, we found significant positive publication bias present; i.e. abstracts with positive results were significantly more likely to proceed to publication than those with negative results (Odds Ratio: 2.01; 95% Confidence Intervals: 1.52 – 2.66; P < 0.001). We then undertook a review of all RCT abstracts presented at the ASA annual meetings between 2010 to 2016. In Chapter 3, we looked to see whether the introduction of mandatory prospective trial registration in 2005 had been followed by a decrease in publication bias in the anaesthesia literature, by comparing the ASA conference abstracts in the first period prior to the introduction of mandatory trial registration (2001-2004) with a period after the introduction of mandatory trial registration (2010-2016). When comparing the odds ratios for abstracts with positive results proceeding to publication over those with negative results, between the two discrete periods of time, we found a marked decrease in the amount of positive publication bias present (P = 0.021) after the introduction of mandatory trial registration. However, after adjustment for study size and abstract quality, a significant amount of publication bias still remained in the anaesthesia literature in the period post introduction of mandatory trial registration (Odds Ratio: 1.34; 95% Confidence Intervals: 1.02 – 1.76; P = 0.037). The results in Chapter 3 contrasted surprisingly to our findings in Chapter 4, where we calculated the proportion of abstracts presented at the ASA annual meetings between 2010 to 2016 that had been prospectively registered, and also quantified the proportion of major discrepancies between prospective trial registration entries and corresponding journal publications. Here, we found that an unacceptably low proportion of RCT abstracts (21%) had undergone prospective trial registration, and almost half (48%) had a major discrepancy, being a primary outcome, sample size, or study intervention change, between their registry data and journal publication. These results signified poor compliance with the implementation of mandatory prospective trial registration, and the presence of significant selective outcome reporting in the anaesthesia literature over the period of time examined. We found no significant reduction of positive outcomes when comparing abstracts without and with prospective registration (Odds Ratio = 0.77; 95% Confidence Intervals: 0.56 – 1.06; P = 0.105). However, rates of journal publication for conference abstracts presented at the ASA annual meeting were found to be substantially higher among prospectively registered trials than in unregistered trials (Odds Ratio = 3.82; 95% Confidence Intervals: 2.73 – 5.35; P < 0.001), indicating that this requirement is clearly influencing publication outcomes and may have contributed to the more even distribution of positive versus negative trial findings among those achieving publication. Improved understanding of the components and contributory factors to publication bias and selective outcome reporting is still required, along with the future implementation and maintenance of more effective strategies against these core problems in medicine and medical research. The uncertainty present in conclusions of meta-analyses and summaries of evidence, stemming from sources of error such as publication bias, needs to be better communicated when guiding clinical practice and future trials.